Vapor-liquid-solid and vapor-solid growth of self-catalyzed GaAs nanowires

The physics and technology of semiconductor nanowires (NWs) are currently among the most investigated topics in the field of nanoscience.¹ Most NWs are grown on semiconducting and insulating substrates using metal catalysts - such as Au - to promote one-dimensional growth, and the vapor-liquid-solid (VLS) model^2,3 has been successful in explaining the experimental growth process. Recently, an increasing effort has been devoted by to develop NW growth methods that avoid the use of Au or any other foreign catalytic materials, to avoid contamination from catalyst incorporation during growth.^4–6 

In the case of GaAs, Au-free NW growth has been obtained for selected substrate and growth conditions, with Ga nanoparticles found at the tip of the resulting NWs.^7–14 Using a VLS-like model, NW growth has been associated with the formation of Ga nanoparticles, which promote one-dimensional growth.^7–14 

An important general feature of GaAs NWs obtained by Ga-assisted growth is their prevalent zincblende (ZB) crystal structure,^9,10,13,14 in contrast with the wurtzite (WZ) structure commonly observed during Au-catalyzed growth. When using Au as a catalyst, growth conditions have to be carefully tuned in order to obtain the pure cubic phase.^15,16 The occurrence of the WZ structure in NWs of materials that exhibit ZB structure in the bulk phase has been widely discussed in the literature. The cohesive energy difference between the two crystallographic phases in GaAs was estimated as 24 meV per III-V atoms pair.¹⁷ Recent first-principle calculations predicted that GaAs NWs with diameters up to about 5 nm are more stable in the WZ form than in the ZB form,¹⁸ the surface density of high-energy dangling bonds playing a leading role in stabilizing the WZ structure for small diameters. The calculated cross-over diameter is, however, exceedingly small as compared to the experimental evidence of WZ NWs with diameters as high as 100 nm.

A growth model has been proposed,^19,20 that relates the occurrence of WZ structure in Au-catalysed GaAs NWs to specific conditions at the growth front of the NW, i.e., at the interface between the catalyst nanoparticle and the semiconductor NW body. More specifically, the model predicts that WZ structure is favored when the relative surface energies of the catalyst particle and of the semiconductor material allow nucleation at the vapor-liquid-solid triple phase line (TL), and in conditions of supersaturation of the Au particle with Ga (and possibly As).The model was applied by Cirlin and co-workers¹⁰ to associate the prevalent ZB structure observed in Ga-catalysed GaAs NWs with the suppression of TL nucleation, following the lower surface energy of Ga as compared to Au.

Spirkovska and co-workers⁹ reported on the possibility of growing Ga-catalysed NWs with mixed ZB-WZ regions by changing the flux conditions, but the actual crystal phase sequence varied from wire to wire and could not be controlled. Recently an extensive study has been published by Plissard et al.²¹ on Ga-assisted growth of GaAs NWs on Si by MBE. The authors investigated the role of V/III equivalent beam pressure ratio (BPR) and growth temperature on the NW structure and morphology. They pointed out that the As/Ga BPR controls the tapering of the NWs and the presence of a Ga nanoparticle at the NW tip, and observed that growth in As-rich conditions leads to the consumption of the Ga droplet and to the end of the VLS growth process. They also stated that no NW growth can occur when the Ga nanoparticles at the NW tip are absent or consumed. The authors report also on a complex crystal structure in NWs where the nanoparticle was completely consumed: a defect-free ZB NW body includes a WZ segment close to the tip, together with a multiply twinned transition region, the very tip being ZB again. A similar structure was observed by Jabeen and coworkers⁷ for self-catalysed GaAs NWs growth on a cleaved Si surface. In that paper the authors also reported about the coexistence of NWs with a Ga nanoparticle at the tip, and NWs with a crystalline GaAs tip. Kogstrup et al.¹³ recently demonstrated the growth of twin-planes free ZB GaAs NWs by self-catalyzed MBE on Si. In the paper, a TEM image of a NW displaying an extended WZ region close to the tip and no evidence of Ga nanoparticle was also shown, but the structure and the morphology of the NW tip was not specifically addressed in the paper.

In this paper we show that both ZB and WZ phase can be found in GaAs nanowires grown by Ga assisted VLS processes depending on the effective As/Ga abundance at the growth front, that determines the size and super-saturation of the Ga nanoparticles at the nanowire tip. We also demonstrate, however, that GaAs NWs nucleated following a VLS mechanism assisted by Ga nanoparticles can continue to grow by a vapor-solid (VS) mechanism even after the consumption of the Ga catalyst particle. Important structural changes are found to take place at the NW growth front when growth conditions lead to the consumption of the catalyst.

GaAs NWs were grown by solid-source molecular beam epitaxy (MBE) on Si-treated n-type GaAs(001) and (111)B wafers.¹⁴ Substrates were prepared as follows. After the thermal desorption of the native oxide, 0.3-μm-thick n-type (Si doped, ND = 1 × 10¹⁶ cm^-3) GaAs buffer layers were deposited at 600°C, at a growth rate of 0.28 nm/sec and with a Ga to As BPR of 10-12, as determined by means of an ion gauge positioned at the sample location. After buffer growth, the sample temperature was lowered to 500°C, and a 4 monolayer (ML) thick (1 ML = 6.26 × 1014 at cm-2) Si layer was deposited under As flux in 200 s. Substrates were then exposed to air for 15’ at 300°C, before being introduced again in the experimental system and degassed at 300°C for 30’.

GaAs NWs were grown by exposing the above substrates to Ga and As beams simultaneously. During most of the growth, the Ga flux was fixed at the value used for the buffer growth, while the As flux was varied in order to obtain different BPRs. In selected cases, the Ga flux was halved to achieve more rapidly As-rich growth condition. The growth was terminated in all cases by interrupting simultaneously the Ga and As fluxes and then cooling the sample to room temperature by removing it from the growth chamber.

The NW morphology was characterized by scanning electron microscopy (SEM) using a Zeiss SUPRA40 instrument equipped with a Shottky field-emission gun (SFEG). Selected NWs were transferred to a carbon-coated copper mesh for structural characterization by high-resolution electron microscopy (HRTEM) and annular dark field (ADF) scanning transmission microscopy using a JEOL 2200 FS instrument equipped with a Shottky FEG emitter and operated at 200KeV, with objective aberration of 0.5 mm and a TEM point resolution of 0.19 nm.

In Figure 1 SEM images of representative NWs grown on Si-treated substrates are shown. High magnification images of NW tips are reported in the insets. The NWs in Fig. 1(a) were obtained on GaAs(100) after 30’ of growth with a BPR of 12 at a substrate temperature of 620° C. The majority of the NWs display a Ga nanoparticle at the tip, as confirmed by energy-dispersive spectroscopy mapping (EDS) (not shown). As extensively reported in Ref. 14, these wires arise from a Ga-catalysed growth process, promoted by an ultra-thin Si oxide layer on GaAs (001), that in a wide temperature range, and in the present flux conditions, favors the formation of Ga nanoparticles and the subsequent Ga-assisted NW growth. In Fig. 1(b) and 1(c) examples of NWs obtained on Si-treated GaAs(111)B are shown. In case of Fig. 1(b) the NWs were grown for 30’ in the same nominal conditions as those in Fig. 1(a). No Ga nanoparticles can be seen at the NW tip. The nanostructures in Fig. 1(c) result from 85’ growth at BPR 6, at a substrate temperature of 620° C. Ga nanoparticles are clearly present at the NW tip.

It appears evident from Fig. 1 that different NW morphologies are obtained depending on the substrate orientation and on the growth parameters, and that the presence or absence of a Ga nanoparticle at the NW tip can be controlled by the V/III ratio. In particular, growth on (111)B substrate with a relatively high As/Ga ratio lead to NWs with no detectable Ga nanoparticles and a crystalline tip. Similar NW morphologies were recently obtained by Plissard et al.²¹ by self-catalysed MBE on Si(111). The authors related the absence of Ga nanoparticles to their transformation in solid GaAs taking place in As-rich conditions, and stated that the nanoparticle disappearance corresponds to the end of NW growth.

Evidence of consumption of the Ga nanoparticles occurring during uninterrupted NW growth can be found in the images in Fig. 2, displaying the extremity of an ensemble of GaAs NWs grown on Si-treated GaAs (100), together with high magnification details of their tips. Together with wires with a large Ga nanoparticle at the tip, similar to that in the inset of Fig.1(a), we see NWs where the Ga nanoparticle is clearly shrinking, (see Fig. 2(b) and 2(d)) or seems to have completely disappeared (Fig. 2(c)). The simultaneous presence on the same substrate of NWs at different stages of Ga nanoparticle consumption suggests the presence of local fluctuations in the relative abundance of As and Ga at the growth front, having an impact on NWs tip morphology.

In order to obtain further insight about the evolution of NWs morphology in growth conditions that lead to the consumption of the Ga nanoparticle, we performed growth of different durations, on Si- treated GaAs(111)B, at a substrate temperature of 620°C and with As/Ga BPR of 12, i.e., in the same conditions used for NWs in Fig. 1(a) and 1(b). In Fig. 3(a) and 3(b) the NW morphology after 1 minute and 5 minutes of growth is examined by SEM. A Ga nanoparticle at the NWs tip is evident at the very beginning of the growth (Fig. 3(a)), confirming the capability of the Si-treated substrate to promote the formation of Ga nanoparticles also in the (111)B surface orientation. Nanoparticles corresponding to NWs growing vertically on the substrate have diameter between 30 and 70 nm. A small number of significantly larger nanoparticles can also be seen on the substrate, suggesting a tendency to coalescence of more than one Ga droplets.²² These larger nanoparticles appear to be related to three-dimensional GaAs nanostructures but not to standing NWs. A detailed study of the size distribution of Ga droplets at the beginning of the growth and its possible dependence on growth parameter and/or specific features of the substrate surface goes beyond the scope of the present work.

The nanoparticles at the NW tip disappear as the growth proceeds. After 5 minutes Fig. 3(b)) the nanoparticles cannot be recognized anymore. In Fig. 3(c) we show the average length of the NWs as a function of the growth time, as measured on (111)B-oriented substrates (solid symbols), and on (100) oriented substrate (open symbols), where growth is assisted by the continued presence of Ga nanoparticles.¹⁴ We observe that growth starts at the same rate on the two substrates, but in the case of the (111)B substrate it slows down. By considering the increase in NW length observed after the first 10 minutes, i.e. after the nanoparticles have completely disappeared, we can estimate a growth rate of 1 nm/s. The data suggest that as long as the growth is assisted by Ga nanoparticles the growth rate is high (6.7 nm/sec), while when the Ga nanoparticle is consumed the growth slows down to a rate that is approximately one seventh. However, the data also indicate that the growth continues, even after the Ga nanoparticle has disappeared, at a rate that is approximately four times higher than in case of two dimensional epitaxial growth in the same conditions. The latter result appears even more clearly in the data shown in Fig. 4. NWs were grown on GaAs(111)B in catalyst assisted conditions (standard Ga flux and BPR = 8) for 15 minutes at a substrate temperature of 600° C. Then the Ga flux was halved, to obtain the consumption of the Ga nanoparticle, and the growth was continued for 5 minutes (Fig. 4(a)) or 45 minutes (Fig. 4(b)). In both samples, together with a number of thin and short NWs whose growth appear to be aborted, we observe a majority population of well-defined, longer NWs. Details of the NW tips are shown in the inset, confirming in both cases the absence of Ga nanoparticles. The average NW diameter is 24±2 nm and 55±2 nm, respectively. The average length of these nanowires is 1.0±0.3 μm in the first case, and 1.9±0.2 μm in the second case, indicating that anisotropic growth proceeded also without the Ga nanoparticles, yielding an average increment in the in the NW length of 0.9 μm and in 40’. This corresponds to a rate of 0.37 nm/s. The observed growth rate compares well to the value of 1 nm/s, estimated from the data in Fig. 3(c) (solid symbols) for the growth after the consumption of the nanoparticle, and obtained with a Ga flux double than here. From the data in Fig. 1–4 it appears that different NW tip morphologies correspond to different growth regimes. In order to investigate about the implication of these different regimes on NW crystal structure, a collection of NWs displaying different morphologies at the tip have been investigated by (S)TEM. Images in Figs. 5(a), 5(c), 5(e), and 5(g) are HRTEM images. Relevant diffractograms are reported on the left. Figs. 5(b), 5(d), 5(f), and 5(h) are STEM-ADF images from morphologically similar tips. These images have been reproduced in false colors to enhance detail visibility. Fig. 5(a) and 5(b) show representative cases of NWs with an amorphous Ga nanoparticle at the tip larger than the NW body, typical of Ga-catalyzed growth. In this case the NW body is made of ZB GaAs. Note in particular that in Fig. 5(a) a few crystalline ZB GaAs planes are visible within the amorphous nanoparticle. Fig 5(b) is obtained along the [110] zone axis and shows a constant ADF intensity in the NW central part, rapidly decreasing on the sides. This is coherent with the typical {110} orientation of the NW sidewalls. This sidewall orientation has been found on all the observed NWs, irrespective of the NW tip morphology. Figs. 5(c) and 5(d) show representative cases of wires with an amorphous tip of the same size of the NW body. This morphology, intermediate between the previous one and those in Figs 5(e)–5(g), corresponds to a necessary stage of the nanoparticle consumption process. Here the NW body below the nanoparticle is comprised of WZ GaAs. Figs. 5(e)-5(f) illustrate a case where the amorphous nanoparticle is missing. A small ZB crystalline GaAs tip is observed on top of a WZ GaAs body. Fig. 5(g)-5(h) shows a similar case where the crystalline cubic tip has the same lateral size of the NW body and shows a fully developed faceting. Also here the crystal structure below the ZB GaAs tip is comprised of WZ GaAs. These data reveal that the different tip morphologies, that correspond to different stages in the process of Ga nanoparticle consumption, are characterized by different crystal structure in the region close to the NW growth front. In particular, data suggest that growth in conditions that lead to the shrinking of the Ga nanoparticle is characterized by WZ structure, while growth after the complete disappearance of the Ga nanoparticle gives rise to ZB structure.

Confirmation to this hypothesis can be found in the TEM analysis of NWs for which the BPR has been switched to intentionally change the growth mode. In Figs. 6 and 7 we show TEM results for two representative NWs grown for 15’ in the presence of a Ga nanoparticle and then for 5’ with reduced Ga flux, to induce the consumption of the Ga nanoparticle and continue the growth in the absence of it, as in the experiment in Fig. 4(a). In both cases moving toward the tip we find a segment with WZ structure, that follows the main ZB nanowire body. The length of the WZ region is different in the two wires, being about 10 nm in Fig. 6 and 75 nm in Fig. 7. The transition region is characterized by a high density of twins. The NW tip exhibit in both cases ZB structure. In Fig. 6 the ZB region at the tip has a length of about 100 nm. The HRTEM image in Fig. 6(a), together with the diffractograms obtained in the highlighted regions, demonstrate the twinned ZB structure of the NW tip. The tip exhibits facets with {110} orientation. On the basis of the results in Fig. 5, we infer that the WZ segments observed in both wires corresponds to the portion of NW grown during the progressive shrinking of the Ga nanoparticle, and that this process lasted different times in the two NWs. In Fig. 8(a) representative NW grown in catalyst assisted mode for 15’ and then in catalyst free mode for 45’, as in Fig. 4(b), is shown. A high resolution image of the wire tip is shown in Fig. 8(b), together with the diffractogram highlighting the ZB structure. The tip presents well defined {110} facets. The NW body below the tip also exhibits a pure ZB structure for a length of the order of 1 μm. This can be appreciated also from the alternation of twinned regions with large defect-free regions that is characteristic of ZB NWs.¹⁴ Referring to the results in Fig. 4, this portion of the wire corresponds to growth by VS mechanism after the consumption of the Ga nanoparticle.

By observing the data in Fig. 5 we can define three NW types: NWs with a Ga nanoparticle larger than the NW diameter, NWs with a Ga nanoparticle with the same diameter as the NW body and NWs that do not display a Ga nanoparticle at all, but a pyramidal crystalline tip. Data in Fig. 3(a) show that NWs nucleate as the first type, but exploration of different growth conditions indicates that an increase in the As/Ga BPR may lead to the disappearance of the Ga nanoparticle. From Fig. 2 it is clear that for certain growth parameters all such cases can be found on the same sample, suggesting that their occurrence may depend on local fluctuations in the relative abundance of As and Ga atoms. Such discrepancy from the nominal growth conditions given by the BPR values can be explained as follows. It is well known that GaAs NW growth by MBE is strongly dependent on the diffusion of adatoms from the substrate surface to the growth front.²³ NWs growth takes place here at relatively high temperature, close to the temperature of incongruent evaporation for GaAs (630°C).²⁴ In this conditions Ga and As fluxes impinge on the whole exposed sample surface (substrate and NW sidewalls) but desorption of As occurs much faster than that of Ga. Arsenic remains therefore almost in the vapor phase, while Ga adatoms diffuse on the surface and toward the growth front at the NW tip. The As/Ga abundance ratio therefore depends on all factors that may affect Ga arrival rate at the NW tip, such as local temperature fluctuations²⁵ and different collection areas (e.g. different NW length and/or inclination, as well as shadow effects). All these features can modify the growth conditions at the single NW tip and make the coexistence of NWs with different tip morphology (as in Fig. 2) possible. In this frame we can understand the differences between NWs in Fig. 1(a) and (b), obtained with the same nominal BPR on substrates with (001) or (111)B orientation, respectively. The inclined orientation of the wires on the (100) substrate allows the collection of Ga atoms from the source beam along a non-negligible portion of the NW side-facets; therefore it favors the arrival of a higher number of Ga atoms at the NW tip compared to the case of vertical growth on (111)B substrate.²⁶ This more abundant Ga flux allows the persistence of the Ga nanoparticles during the growth on (100) surface, while on (111)B they are consumed.

Analysis of the data in Figs. 5–8 shows that close to the tip the three types of nanowires are characterized by different crystal structures: NWs of the first type exhibit ZB structure in contact with the Ga nanoparticle (Fig. 5(a) and 5(b)), NWs of the second type exhibit WZ structure in the same region (Fig. 5(c) and 5(d)), while NWs that do not display a clearly defined nanoparticle have a WZ section, but are ZB at the very tip (Fig. 5(e)–5(h), Figs. 6 and 7). First type NW morphology with ZB structure is commonly observed in self-catalysed GaAs NWs grown in Ga-rich conditions.^9,10,13,14 Cirlin et al. interpreted it in terms of the suppression of TL nucleation, as a consequence of the low surface energy of liquid Ga, in condition of low As-supersaturation of the nanoparticle.¹⁰ The TEM data in Fig. 5(a) indeed reveal the presence of few layers of GaAs within the Ga NP, as expected if new monolayers nucleate within the nanoparticle. The consumption of the Ga nanoparticle, observed when the local As/Ga ratio increases and leading to the second type morphology, is accompanied by a decrease of the contact angle between the nanoparticle and the NW (see Fig. 5(b)), and by an increase of As super-saturation of the Ga nanoparticle. According to Refs. 10, 19, and 20 both facts point toward the stabilization of the WZ structure, which we indeed observe in with the region below the Ga nanoparticle in NWs of the second type. In this framework, the WZ segments observed in otherwise ZB NWs in Figs. 5(c), 6, and 7 witness transient phases of growth in As-supersaturated condition, occurring during the consumption of the Ga nanoparticles. The length of the WZ segments, as well as their position relative to the tip, vary from nanowire to nanowire (Figs. 6 and 7), as a consequence of the variation in the local growth conditions during the same growth.

In this work WZ structure has only been observed at the tip of NWs with small Ga nanoparticles, compared with the NW radius, or sandwiched between ZB regions in NWs with crystalline tip, in both cases as signature of growth occurred in a transition phase when the local Ga/As ratio became too low to maintain a stable large size Ga nanoparticle. Nevertheless, the observation of untapered, defect-free WZ region of different length in different growth experiments (up to 500 nm, not shown) lead us to speculate that a careful tuning of the growth condition, involving the control of NW orientation and density, should allow to define growth protocols for steady state growth in the WZ structure. However this goes beyond the scope of the present work.

Data in Figs. 3 and 4 demonstrate that highly anisotropic GaAs NWs growth can continue without the presence of a catalytic nanoparticle by a vapor-solid (VS) mechanism. The growth rate associated to this process, despite significantly lower that the Ga- assisted, is about four times higher than the two dimensional homoepitaxial growth rate in the same flux conditions, demonstrating the presence of a mechanism that favors the incorporation of adatoms at the NW tip also in the absence of a catalyst nanoparticle. The portions of NWs grown by VS mechanism after catalyst consumption, again have ZB structure, as shown in Figs. 6–8. WZ GaAs has been observed only in catalyst-assisted NWs or in bulk material under pressure.²⁷ The ZB structure of the VS-grown NWs can therefore be understood as due to the restoration of the stable GaAs structure in the catalyst-free growth process.

The microscopic mechanism leading to the anisotropic VS growth, occurring significantly faster at the NW tip than on the NW lateral facets, remains an open point. VS growth, in the absence of a growth catalyst, is well known to take place in a wide range of material systems, ranging from metals to oxides and compound semiconductors. The anisotropic growth has been explained either by the presence of a screw dislocation along the growth axis, giving favorable incorporation sites at the tip,^28,29 or by the mobility differences of adatoms on the crystal planes forming the nanowire walls or the tip.³⁰ 

Concerning compound semiconductors, NW growth in the absence of catalyst nanoparticles is routinely obtained for III-N materials. Growth of Ni-catalyzed and catalyst free GaN NWs in otherwise same conditions was found to take place at very different growth rates,³¹ highlighting the enhancement of material incorporation at the nanowires tip related to the presence of the metal nanoparticle, as found in the present work. The GaN VS growth mechanism was recently described in terms of a sticking coefficient on the WZ (0001) c-plane tip higher than on the lateral {1100} m-planes walls.³² 

In the present case we can rule out the systematic presence of screw dislocations along the wire axis, that would be visible by high resolution ADF.³³ On the other hand, the tip facets, as observed after the growth, appear to belong to the {110} family, therefore equivalent to the sidewalls. A different mechanism should therefore be devised, and this will be the object of future investigations.

In this work we explored different growth conditions during self-catalyzed synthesis of GaAs nanowires on Si-treated GaAs by molecular beam epitaxy. From the experimental systematics we determined how the effective As/Ga local abundance ratio and the corresponding nanowire tip morphology influences the nanowire growth rate and final crystal structure. In particular, we found that GaAs NWs can grow by vapor-liquid-solid processes in both zincblende and wurtzite phase depending on the effective As/Ga ratio that determines the size and supersaturation of the Ga nanoparticles at the nanowire tip.

We also demonstrate that a high effective As/Ga beam equivalent pressure ratio can lead to the complete consumption of the Ga catalyst nanoparticle, but that GaAs nanowires can continue to grow by a vapor-solid mechanism that results in a zincblende structure for the nanowire.

The transition from catalyst-assisted to catalyst-free growth is accompanied by a change in crystal symmetry, characterized by the presence of the wurtzite structure during the consumption process of the Ga nanoparticle.

This research was partially funded by Commissariato del Governo di Trieste through Fondo Trieste and by Regione Friuli Venezia Giulia L.R. 47/78-1953 Ambient and Biological Sensors.